<div dir="ltr"><div><div><div>Christoph,<br><br></div> So it does appear that using a more diffusive solver corrected the issue. If I use the HLL solver (this appears to be a very diffusive solver!) it will run properly. I'm in the process of sorting out how to move back to less diffusive solvers, but in the meantime this does correct the issue.<br><br></div><div> As for time stepping, I have explicit subcycling in to handle the timesteps for the cooling, and optionally can switch to an implicit solver if needed. So hopefully that part is okay.<br></div><div><br></div>Thanks so much for the help, I'm currently trying to read up on the different solvers in Toro's book along with ZIngale's online book at astro bookshelf.<br><br></div><div>Warm regards,<br></div><div><br></div>Josh<br></div><br><div class="gmail_quote"><div dir="ltr">On Fri, Jul 3, 2015 at 2:53 AM Christoph Federrath <<a href="mailto:christoph.federrath@gmail.com">christoph.federrath@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><br>
Hi Joshua,<br>
<br>
this is probably related to hydro and EOS and might be a problem of time stepping. Have you modified the timestep constraints in your implementation of cooling? Your cooling has a timescale and if the timestep is too large for a cooling that is too strong and too fast, then this might cause some hydro/EOS inconsistency, which can manifest in negative density/pressure (or negative internal energy, as seems to be the case in your warning below; the warning comes from GC filling, but that the problem arises in hydro/eos). You could try to reduce the CFL number, but eventually, you’d probably have to implement a timestep limiter to properly account for your cooling.<br>
<br>
Cheers,<br>
<br>
Christoph<br>
<br>
PS: Plus, some hydro/MHD solvers might be more stable for a bit longer, but I given that you tested a few and it always eventually failed when the density get high (and thus the cooling get strong), there is clearly a more fundamental problem (likely the one I describe above), which cannot be fixed by just switching to a different hydro/MHD solver.<br>
<br>
<br>
On 30 Jun 2015, at 21:46, Joshua Wall <<a href="mailto:joshua.e.wall@gmail.com" target="_blank">joshua.e.wall@gmail.com</a>> wrote:<br>
<br>
> Dear Flash users,<br>
><br>
> I'm currently testing a cooling function I've written using the SinkRotatingCloudCore collapse problem from the provided simulations. I've set the gamma=5/3 and swtiched to the ideal EOS. My cooling function seems to be working correctly in testing outside of Flash, and cools properly for a given cell at a certain temperature and density.<br>
><br>
> However when I try and let this problem form a sink particle, the cloud collapses and refines down to a high density, then begins to give the gr_sanitizeDataAfterInterpolation warnings like:<br>
><br>
> WARNING after gc filling: min. unk(EINT_VAR)=-2.4992047803499256E+82 PE=3 block=269 type=2<br>
><br>
> This seems to happen close to sink formation, and usually (maybe always, but I can't remember all of the runs!) after a refinement. I've tried setting gr_sanitizeDataMode = 3, but this only delays the issue. Also, if I increase the minimum temperature for the cooling, it delays the issue, but it always still occurs right up until temperatures that prevent the cloud from collapsing. I've also tested by switching off sink particles, and it still occurs, so its not sink formation.<br>
><br>
> I saw in the flash users list history some mention of whether to apply sourceTerms to all the blocks or just leaf blocks, but it now appears to me in Driver_evolveFlash that all blocks are evolved, so I'm not sure if there's a fix there.<br>
><br>
> I've also tested this with the split solver and the unsplit Roe and PPM solvers, and see the same behavior.<br>
><br>
> Any suggestions would be most welcome and appreciated at this stage, if anyone has any ideas of how I might tackle this problem.<br>
><br>
> King regards,<br>
><br>
> Joshua Wall<br>
> Doctorial Candidate<br>
> Drexel University<br>
<br>
</blockquote></div>